Is email revenue incremental, or would those people have bought anyway

What an attributed number is actually claiming

Attribution says an order happened after somebody opened or clicked, inside a window the platform allows. That's a statement about sequence, not about cause. Cause is a different claim, and no email platform makes it, however the dashboard is worded.

Fine, until somebody spends money on the strength of it. A loyal customer who was going to reorder on Friday opens Thursday's email, orders on Friday, and lands in the email column. Nothing about that sale was created by the email.

None of that is the platform being dishonest. Opens, clicks and orders are what it can see, and sequence is the only relationship it can prove between them. The dishonesty, where there's any, happens in the retelling: when a column heading turns into a claim about what would have happened otherwise.

The window decides most of the number

The attribution window is a setting. Widen it and the email column grows without a single extra sale in the business. That's how two agencies reporting the same month hand you two different numbers and both tell the truth.

So read the unattributed part of that chart as well. Total business revenue is the number your bank agrees with, and a programme that's genuinely adding sales moves that one too.

The only clean answer is a holdout

Hold a random slice of the audience out of a send, then compare what both groups spent over the same window. The difference is the closest thing to an honest answer you'll get without hiring a statistician.

1. Pick one repeating send: a campaign to a stable segment, or a single flow
2. Split the audience at random, not by engagement, or you've compared two different kinds of people
3. Hold one group out and leave everything else about the period alone
4. Compare revenue per recipient across both groups over the same number of days
5. Repeat it, because one round on a small list tells you about luck, not about email
6. Write down what you expected before you look, so the result can't be reinterpreted afterwards

Two warnings, both learned the expensive way. Holding people out of an automation that works costs you real orders while you learn, so start with the sends you already suspect. And never run one across a peak trading week. Everything else in the business is moving then too.

Then read the answer properly. A small gap tells you the send is worth keeping and not worth expanding. Almost no gap tells you more than any subject line test you'll run this year.

The emails most likely to be adding sales

• Winback and lapsed customer sends, where the alternative outcome is genuinely nothing
• Replenishment timed to the point somebody runs out, which moves a purchase that might otherwise go elsewhere
• Checkout and browse recovery, where the session died and nobody was coming back to it
• Back in stock, because the customer told you they wanted something they couldn't have
• First order welcome sends to people who have never bought, where there's no habit to ride on

See the pattern? Incremental email reaches somebody who wasn't already on their way to the till. That's why lifecycle strategy is a better lens for this than a campaign calendar. It's built around the moments where the answer would otherwise have been no.

Discounting people who were already at the till

The least incremental email in ecommerce is a discount sent to people who were about to buy at full price. It reports beautifully. It also hands your margin away, and the report has no way of showing you that.

Same goes for a reminder sent to somebody with an item in the basket and a card in their hand. Some of those sales get recovered. Some were never at risk, and only a holdout tells you which is which.

There's a version of this that never shows up in the revenue line at all. A code used by somebody who was buying anyway is a discount on a sale you already had. Enough of those and a programme reports growth while the business earns less on every order.

What we report, and how to argue with it

Our case studies publish attributed revenue, because that's what the platform measures and inventing a better number would be worse. We publish the period it covers and the account it came from too, so you can push back on it.

Ask any agency two questions before you believe a figure on their site. What window was it measured over, and what did total business revenue do in the same period. No answers, no number: it's decoration. The buyer's guide has the rest of that conversation.